Bioinformatiics data processing power depends on CPU L3 cache A LOT!

When speccing a CPU for a new bioinformatics computer we tend to focus on threads and frequency, but do you look at the cache? Well you definitely should. On January 30, I started jobs on two servers. The first one has 2x Xeon E5-2680 v3 (48 threads total) and the second has 1x Xeon E5-2667 v3 (16 threads). The work we are doing is processing raw RNA-seq data using a pipeline of Skewer+STAR+Kallisto. Server 1 has many more threads, so I ran two parallel jobs of 12 threads each, while on Server 2 I'm running one job using 8 cores. The cluster nodes are attached to the same network storage, so I/O capacity is the same. Since Jan 30, these two machines have collectively processed 4670 datasets and the difference between these two machines was a big surprise. Not only did Server 2 process more datasets, when normalised by number of threads, Server 2 processed 4.1 times more datasets! 

Server 1Server 2
CPU2x Xeon E5-2680 v3
1x Xeon E5-2667 v3
Cores2416
Threads4816
Frequency (GHz)3.33.2
Cache L1 (kB)768512
Cache L2 (MB)64
Cache L3 (MB)60320
DEE2 parallel jobs21
DEE2 cores248
Completed jobs19732697
Jobs per thread82.21337.13

These CPUs are from the same generation and run at a similar frequency, but the main point of difference is with the massive L3 cache of Server 2.

I was a bit confused with how Xeon E5-2667 v3 could have such a huge cache, as the online documentation indicates only a 20 MB L3 cache, but noticed that the lscpu command indicated this server had 16 sockets (!), so perhaps this is a special CPU that has 16 times the normal cache. I wasn't around when this cluster was set up but it certainly gives us an insight into how important cache is for CPU performance. So does increased cache lead to a linear increase in processing power? Well here Server 2 has 5.3 times more L3 cache, and the throughput was 4.1 more, so yes the increase is nearly linear and would be a worthwhile to spend more for a bigger L3 cache! 

With the trend of increasing L3 cache on AMD X3D chips (up to 128 MB), it would be great to benchmark the per-thread performance for bioinformatics tasks on those Ryzen CPUs compared to the low cache non-X3D chips. In the lack of actual benchmarks we can guess that double the cache could yield 40-50% more performance, but given the 44% premium being charged for the extra cache, the extra cost may only be justified if the machine will be heavily used.

Ryzen 7950X (64 MB L3 cache) AUD$829

Ryzen 7950X3D (128 MB L3 cache) AUD$1199

Source: Scorptec (13 Feb 2025)






Location: Melbourne VIC, Australia

Popular posts from this blog

Data analysis step 8: Pathway analysis with GSEA

Image
In our RNA-seq series so far we've performed differential analysis and generated some pretty graphs, showing thousands of differentially expressed genes after azacitidine treatment. In order to understand the biology underlying the differential gene expression profile, we need to perform pathway analysis. We use Gene Set Enrichment Analysis ( GSEA ) because it can detect pathway changes more sensitively and robustly than some methods. A 2013 paper compared a bunch of gene set analyses software with microarrays and is worth a look. Generate a rank file The rank file is a list of detected genes and a rank metric score. At the top of the list are genes with the "strongest" up-regulation, at the bottom of the list are the genes with the "strongest" down-regulation and the genes not changing are in the middle. The metric score I like to use is the sign of the fold change multiplied by the inverse of the p-value, although there may be better methods out there...
Read more

Uploading data to GEO - which method is faster?

If you have had to upload omics data to GEO before, you'll know it's a bit of a hassle and takes a long time. There are a few methods suggested by the GEO team if you are using the Unix command line: Using 'ncftp' ncftp set passive on set so-bufsize 33554432 open  ftp://geoftp:yourpasscode@ftp-private.ncbi.nlm.nih.gov cd uploads/your @mail.com_ yourfolder put -R Folder_with_submission_files Using 'lftp' lftp  ftp://geoftp:yourpasscode@ftp-private.ncbi.nlm.nih.gov cd uploads/ your @mail.com _ yourfolder mirror -R Folder_with_submission_files Using 'sftp' (expect slower transfer speeds since this method encrypts on-the-fly) sftp  geoftp @s ftp-private.ncbi.nlm.nih.gov password: yourpasscode cd uploads/ your @mail.com _ yourfolder mkdir new_geo_submission cd new_geo_submission put file_name Using 'ncftpput' (transfers from the command-line without entering an interactive shell) Usage example: ncftpput -F -R -z -u  geoftp  -p "yourpasscode...
Read more

Generating a custom gmt file for gene set analysis

Image
Pathway and gene set analysis is a common procedure for interpretation of RNA-seq or other genome-wide expression assays. Most of the time, we use GSEA to tell us whether our gene sets of interest are up- or down-regulated. We can use gene sets from KEGG , Reactome , GO ,  MSigDB  and other sources, but you can also generate your own gene sets. The format used for GSEA is gmt . I'm going to take you through two examples of generating custom gene sets: Generate gene sets from published data sets using GEO2R Let's say you're interested in the transcription factor STAT1. I found a dataset in GEO called "Knockdown of STAT1 in SCC61 tumor xenografts leads to alterations in the expression of energy metabolic pathways", which has a paper in BMC Med.  Most uploaded array data sets can be reanalysed with GEO2R, which runs the array analysis tool Limma but this is embedded in the webpage and has a GUI which makes it very accessible for biologists. Click this link ...
Read more